Systems · Embedded Linux · Edge AI
I turn messy real-world signals into decisions on constrained hardware.
28 years of it — from Manchester encoding in software to camera-radar fusion at the edge.
// Zero-copy: a borrowed view over a camera frame.pub fn detect(frame: &FrameView<'_>) -> Vec<Box2D> { let tensor = preprocess(frame); // no allocation on the hot path let raw = engine.infer(&tensor); decode(raw, frame.width, frame.height)}I've been working with Linux since 1998 — 28 years and counting — with a focus on embedded systems. Along the way I've built full protocol stacks down to the media access layer (power line, wireless, Ethernet), early smart-grid metering and control on microcontrollers, industrial machine monitoring for predictive maintenance, Linux driver and BSP work across many platforms, and computer vision pipelines. I've been at Au-Zone Technologies since 2011, now VP of R&D, working on spatial perception at the edge: computer vision, fused with lidar and radar down to the raw radar cube, to detect and localize things and stuff on sub-10W devices.
Latest writing
All posts →First posts are on the way.
Speaking
All talks →Technical Presentation · 2026
From YOLO to SAM: Segmentation Models on Real Edge Hardware
Embedded Vision Summit